Can increased testing produce fewer infections?

Julio Montaner and his team provided further evidence to strengthen the case that antiretroviral coverage was bringing down the rate of new diagnoses in a subsequent paper published¹ in The Lancet.

British Columbia (BC) provides a good test case for treatment as prevention, as very little of the data needed goes missing. One single provincial public health reference laboratory does more than 90% of all HIV testing in British Columbia and provides surveillance data on new diagnoses to the provincial Center for Disease Control. Moreover, universally available antiretroviral therapy is provided by one single centre, the British Columbia Centre for Excellence in HIV/AIDS, which collects data on ARV use, CD4 count and viral load.

In his Lancet piece, Montaner supplies a graphic showing that as the number of people on ARVs rose, the number of new HIV diagnoses per year sank, and that the two curves closely mirrored each other. He also shows that the falls in diagnosis and rises in ARV coverage seen in 1996-99 and again in 2005-9 were unlikely to be chance events and that the majority of the fall in diagnoses, especially in 2007-9, was in injecting drug users.

His study, however, because it is an ecological study, cannot show causation; it cannot prove beyond doubt that the fall in diagnoses was caused by the increased uptake of ARVs, which is Montaner’s hypothesis. It is possible that the fall in diagnoses is, instead, caused mainly or partly by other factors, such as serosorting (see Serosorting, sexual harm reduction and disclosure), the improved efficacy, rather than the improved uptake, of ARVs, or by other behavioural, social or medical influences.

He also does not, in this diagram, show the ‘beginning and the end’ of the process. He does not include data, firstly, on whether the fall in diagnoses is associated with an increase in testing, nor does he show the second part of the presumed link between more testing and fewer diagnoses, namely an increase in the proportion of people with an undetectable viral load. However, he does provide the data which can be added in, producing the following graphic:

In this diagram, we see that the curve for the proportion of people with viral loads under 50 closely parallels the number on ARVs in general, especially in non-IDUs (though IDUs catch up later). Still more significantly, we also see that from 2003 onwards, the proportion of people with an undetectable viral load is paralleled quite closely by the increase in the number of HIV tests done.

Montaner did show that the increase in the proportion of people in general (p=0.001) and IDUs (p=0.002) with viral loads under 50 was highly statistically significant.

He was also able to calculate that for every 100 additional individuals taking ARVs, the estimated number of new HIV cases decreased by 3%, and for every tenfold decrease in viral load, the number of new HIV cases decreased by 14% (though with a wide confidence interval of 2 to 25%).

Finally, he also showed that the decrease in HIV diagnoses occurred at a time of increases in bacterial STIs, with a 150% increase in the annual incidence of gonorrhoea between 1996 and 2008 (from 0.0126 to 0.0313 per 100 population), thus making it unlikely that the reduction in HIV diagnoses seen was due to increased safer sex in the conventional sense (i.e. abstinence, faithfulness or condom use).